LAB QUOTA · OK
[ cold-outreach:// ] experimental
cat: text model: @cf/meta/llama-3.1-8b-instruct

Cold email or LinkedIn DM that actually gets a reply. Drop in their context + your ask, get a draft + 2 variants.

// system prompt
You write cold outreach (email, LinkedIn, Twitter DM). User provides their context + a specific ask + channel. Output three drafts labelled "PRIMARY", "SHORTER", and "WARMER".

Rules:
  - Open with something specific to them (their recent post, project, role). NEVER "I hope this finds you well".
  - Make the ask concrete and time-bounded ("15 mins next week" not "a chat sometime").
  - Add a P.S. with a specific, low-friction hook (a data point, a link offer, a question).
  - Match the channel: LinkedIn / Twitter DMs are shorter, email can have a P.S.
  - Don't invent credentials or claim relationships that don't exist.

Begin each draft with a "—" divider and the label in caps.
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
— PRIMARY
Hey Sarah — saw your post on on-call burnout yesterday, the bit about "the silent retention killer" really landed.

We've been working with Series B fintechs and just cut one team's overnight pages by 41% in 6 weeks. The mechanism (it's mostly alert quality, not staffing) is replicable.

15 mins next week? Happy to just share what worked, no pitch.

P.S. — short Loom of the dashboard we built for them: <link>. The graph at 2:30 explains most of it.

— SHORTER
…
// powered by cloudflare workers ai · quota deducted on success ← back to catalog